IXN 1 — Exploratory Data Analysis

#IMPORT (wrangled) data 
df_raw <- read_csv("data/CLEAN_UtterancesForAnalysis.csv") 

#WRANGLE into DF of coded utterances 
#NOT unique utterances, 1 obs for each utterance+detail-code
df_coded <- df_raw %>% 
  #rename and factorize cols
  mutate(
    #UNIQUE IDS
    sid = factor(SID), #unique ID for utterance+detail-code
    pid = factor(PID, levels = c( #define level order so happiness first
      #HAPPINESS-FIRST    
      "bjs827ee1u", "3r2sh20ei", "4728sjuiz","7ACC0B75","92ghd48xe","iurmer289", "s294hoei",
      #SPACE-FIRST    
      "j2719eertu2","lkin27js09b","li832lin23","7382kwtue","E1D39056","8v892iige")),   
    #create unique ID for utterances
    uid = factor(as.numeric(factor(paste(pid,factor(Utterance))))), #construct a unique ID for utterances
    #recode lower case and order based on true task order
    TASK = factor(recode(Condition, "Static"="static", "Interactive"="ixn" )),
    TASK = factor(TASK, levels = c("static", "ixn")), #reorder factor levels
    #rename Notebook as DATASET
    DATASET = factor(recode(Notebook, "Happiness"="happiness", "Space"="space")),
    #create temp dataset order var
    data_order = factor(paste(TASK,"_",DATASET)), #create an order var 
    data_order = recode(data_order, "ixn _ happiness"="space-first",
                                    "ixn _ space"="happiness-first",
                                    "static _ happiness"="happiness-first",
                                    "static _ space"="space-first"),
    utterance = Utterance,
    reps_group = factor(`group`),
    reps_all = factor(`All representations`),
    #rename flags
    flag_story = `Flag Storytelling`,
    flag_correction = `Flag Correction`,
    flag_simultaneous = `Flag Simultaneous Characterization`,
    #recode and order TOP LEVEL CODES 
    code_topic = factor(Highlevel),
    code_topic = recode(code_topic, "ANALYSIS PROCESS" = "PROCESS"),
    code_topic = factor(code_topic, levels = c("PROCESS","DATASET","VARIABLE","RELATIONSHIP")),
    code_datatype = factor(`Data Type`),
    code_detail = factor(`Utterance Type`),
    timestamp = adj_timestamp,
    ixn = factor(interaction_used), #was interaction used?
    PNUM = factor(PNUM,levels = c("P6", "P9", "P10", "P2", "P4", "P12","P13", 
                                   "P5", "P7", "P8", "P3", "P1","P11")),
    
    ) %>% 
  select(sid,pid,PNUM,uid,TASK,DATASET,timestamp,ixn,code_topic,code_detail,code_datatype,
         flag_story, flag_correction, flag_simultaneous, utterance, reps_group, reps_all, data_order) %>% 
  arrange(data_order)

#REPLACE NA in logicals to FALSE  
df_coded$flag_story[is.na(df_coded$flag_story)] <- FALSE
df_coded$flag_correction[is.na(df_coded$flag_correction)] <- FALSE
df_coded$flag_simultaneous[is.na(df_coded$flag_simultaneous)] <- FALSE



#CALCULATE RELATIVE TASK TIMES
df_time <- df_coded %>% mutate(
  time = hms::as_hms(timestamp)
) %>% group_by(pid, TASK) %>% 
  # dplyr::summarise( .groups="keep",
  mutate(
    task_start = hms::as_hms(min(time)),
    task_end = hms::as_hms(max(time)),
    task_mins = round(difftime(task_end,task_start, units="mins"),1),
    task_second = task_end - task_start,
    relative_time = timestamp-task_start,
    rel_time = as.double(relative_time)
  ) %>% ungroup() %>% select(pid,PNUM, code_topic,code_detail, TASK,DATASET,timestamp,task_start,relative_time,rel_time)

**There are 742 rows in the df_coded dataset, where each row represents an utterance coding (i.e. utterance + detail code). There are 662 unique utterances. The difference indicates utterances that were dual-coded (i.e. two detail-level codes). No more than two codes were applied to a single utterance. For the purposes of analysis, dual-coded utterances will be treated as two utterances, as they have two distinct (but lexically insepeperable) units of meaning.

1 DATA PROFILE

df_coded%>% summarytools::dfSummary(
             plain.ascii  = FALSE,
             graph.magnif = 0.75,
             style        = "grid",
             tmp.img.dir  = "temp",
             missing.col = FALSE, 
             method = "render"
)

1.0.1 Data Frame Summary

1.0.1.1 df_coded

Dimensions: 742 x 18
Duplicates: 0

No	Variable	Stats / Values	Freqs (% of Valid)	Graph	Valid	Missing
1	sid [factor]	1. 0 2. 1 3. 2 4. 3 5. 4 6. 5 7. 6 8. 7 9. 8 10. 9 [ 732 others ]	1 ( 0.1%) 1 ( 0.1%) 1 ( 0.1%) 1 ( 0.1%) 1 ( 0.1%) 1 ( 0.1%) 1 ( 0.1%) 1 ( 0.1%) 1 ( 0.1%) 1 ( 0.1%) 732 (98.7%)		742 (100.0%)	0 (0.0%)
2	pid [factor]	1. bjs827ee1u 2. 3r2sh20ei 3. 4728sjuiz 4. 7ACC0B75 5. 92ghd48xe 6. iurmer289 7. s294hoei 8. j2719eertu2 9. lkin27js09b 10. li832lin23 [ 3 others ]	29 ( 3.9%) 103 (13.9%) 43 ( 5.8%) 28 ( 3.8%) 56 ( 7.5%) 87 (11.7%) 88 (11.9%) 82 (11.1%) 48 ( 6.5%) 51 ( 6.9%) 127 (17.1%)		742 (100.0%)	0 (0.0%)
3	PNUM [factor]	1. P6 2. P9 3. P10 4. P2 5. P4 6. P12 7. P13 8. P5 9. P7 10. P8 [ 3 others ]	29 ( 3.9%) 103 (13.9%) 43 ( 5.8%) 28 ( 3.8%) 56 ( 7.5%) 87 (11.7%) 88 (11.9%) 82 (11.1%) 48 ( 6.5%) 51 ( 6.9%) 127 (17.1%)		742 (100.0%)	0 (0.0%)
4	uid [factor]	1. 1 2. 2 3. 3 4. 4 5. 5 6. 6 7. 7 8. 8 9. 9 10. 10 [ 652 others ]	2 ( 0.3%) 1 ( 0.1%) 1 ( 0.1%) 1 ( 0.1%) 1 ( 0.1%) 1 ( 0.1%) 1 ( 0.1%) 2 ( 0.3%) 1 ( 0.1%) 1 ( 0.1%) 730 (98.4%)		742 (100.0%)	0 (0.0%)
5	TASK [factor]	1. static 2. ixn	396 (53.4%) 346 (46.6%)		742 (100.0%)	0 (0.0%)
6	DATASET [factor]	1. happiness 2. space	424 (57.1%) 318 (42.9%)		742 (100.0%)	0 (0.0%)
7	timestamp [hms, difftime]	min : 622 med : 2857 max : 6900 units : secs	622 distinct values		742 (100.0%)	0 (0.0%)
8	ixn [factor]	1. FALSE 2. TRUE	633 (85.3%) 109 (14.7%)		742 (100.0%)	0 (0.0%)
9	code_topic [factor]	1. PROCESS 2. DATASET 3. VARIABLE 4. RELATIONSHIP	160 (21.6%) 176 (23.7%) 122 (16.4%) 284 (38.3%)		742 (100.0%)	0 (0.0%)
10	code_detail [factor]	1. data orientation 2. data provenance 3. data size 4. distribution outlier (var 5. distribution range [min, 6. distribution shape [shape 7. distribution variance (sd 8. missing data 9. outlier (relationship) 10. plan of action [ 8 others ]	16 ( 2.2%) 11 ( 1.5%) 9 ( 1.2%) 9 ( 1.2%) 33 ( 4.4%) 79 (10.6%) 1 ( 0.1%) 76 (10.2%) 20 ( 2.7%) 52 ( 7.0%) 436 (58.8%)		742 (100.0%)	0 (0.0%)
11	code_datatype [factor]	1. distribution (continuous 2. distribution (categorical 3. relationship (categorical 4. relationship (categorical 5. relationship (continuous 6. relationship (multivariat	76 (17.8%) 54 (12.7%) 28 ( 6.6%) 55 (12.9%) 146 (34.3%) 67 (15.7%)		426 (57.4%)	316 (42.6%)
12	flag_story [logical]	1. FALSE 2. TRUE	700 (94.3%) 42 ( 5.7%)		742 (100.0%)	0 (0.0%)
13	flag_correction [logical]	1. FALSE 2. TRUE	733 (98.8%) 9 ( 1.2%)		742 (100.0%)	0 (0.0%)
14	flag_simultaneous [logical]	1. FALSE 2. TRUE	682 (91.9%) 60 ( 8.1%)		742 (100.0%)	0 (0.0%)
15	utterance [character]	1. [Talking about the profil 2. actually, let me see if p 3. Although we have like les 4. And are they within range 5. And confidence in governm 6. And just I want to see ho 7. NA 8. NA 9. Because it does seem like 10. Data frame. Got a bunch o [ 652 others ]	2 ( 0.3%) 2 ( 0.3%) 2 ( 0.3%) 2 ( 0.3%) 2 ( 0.3%) 2 ( 0.3%) 2 ( 0.3%) 2 ( 0.3%) 2 ( 0.3%) 2 ( 0.3%) 722 (97.3%)		742 (100.0%)	0 (0.0%)
16	reps_group [factor]	1. barplot 2. columns 3. columns_data_dictionary 4. data_dictionary 5. data_dictionary_dataframe 6. data_dictionary_describe 7. dataframe 8. dataframe_describe 9. dataframe_heatmap 10. dataframe_pairplot [ 15 others ]	16 ( 2.2%) 4 ( 0.5%) 1 ( 0.1%) 59 ( 8.0%) 1 ( 0.1%) 10 ( 1.3%) 80 (10.8%) 1 ( 0.1%) 1 ( 0.1%) 1 ( 0.1%) 568 (76.5%)		742 (100.0%)	0 (0.0%)
17	reps_all [factor]	1. affect_corruption_brush_7 2. Age_CryoSleep_scatterplot 3. age_CryoSleep_ShoppingMal 4. Age_RoomService_scatterpl 5. age_roomservice_scatterpl 6. Age_RoomService_scatterpl 7. Age_ShoppingMall_scatterp 8. altair_profile_contVars_j 9. alx_barplot_df_homeplanet 10. alx_barplot_df_homeplanet [ 245 others ]	4 ( 0.6%) 3 ( 0.4%) 1 ( 0.1%) 1 ( 0.1%) 1 ( 0.1%) 1 ( 0.1%) 1 ( 0.1%) 1 ( 0.1%) 2 ( 0.3%) 1 ( 0.1%) 693 (97.7%)		709 (95.6%)	33 (4.4%)
18	data_order [factor]	1. space-first 2. happiness-first	308 (41.5%) 434 (58.5%)		742 (100.0%)	0 (0.0%)

2 UTTERANCES

2.1 by TASK

print("BY TASK")

[1] “BY TASK”

freq(df_coded$TASK, 
     cumul      = FALSE,
     headings   = FALSE,
     report.nas = FALSE,
     plain.ascii = FALSE) 

	Freq	%
static	396	53.37
ixn	346	46.63
Total	742	100.00

2.2 by TASK and DATASET

#COUNT BY TASK AND DATASET
ctable(x = df_coded$TASK, 
       y = df_coded$DATASET, 
       prop = "t")  

Cross-Tabulation, Total Proportions
TASK * DATASET
Data Frame: df_coded

	DATASET	happiness	space	Total
TASK
static		256 (34.5%)	140 (18.9%)	396 ( 53.4%)
ixn		168 (22.6%)	178 (24.0%)	346 ( 46.6%)
Total		424 (57.1%)	318 (42.9%)	742 (100.0%)

#DF SUMMARIZED BY TASK + DATASET
df_summary <- df_coded %>% 
  group_by(TASK,DATASET) %>% 
  dplyr::summarise(
    c = n()
  )

#STACKED BAR BY TASK
ggplot(df_summary, aes(x = TASK, y=c, fill= DATASET)) + 
  geom_col() + 
  geom_text(aes(label=c), size = 3, hjust = 0.5, vjust = 1.5, position = "stack") + 
  # scale_fill_brewer(type="qual", palette = 4) +
  labs( title = "Utterances by TASK and DATASET",
        subtitle = "More utterances in STATIC; more utterances in HAPPINESS",
        x= "TASK", y = "count") + theme_minimal() 

# + theme(legend.position = "blank")

2.3 by PARTICIPANT

#COUNT BY PARTICIPANT AND TASK
ctable(x = df_coded$PNUM, 
       y = df_coded$TASK, 
       prop = "r")  

Cross-Tabulation, Row Proportions
PNUM * TASK
Data Frame: df_coded

	TASK	static	ixn	Total
PNUM
P6		11 (37.9%)	18 (62.1%)	29 (100.0%)
P9		63 (61.2%)	40 (38.8%)	103 (100.0%)
P10		30 (69.8%)	13 (30.2%)	43 (100.0%)
P2		18 (64.3%)	10 (35.7%)	28 (100.0%)
P4		28 (50.0%)	28 (50.0%)	56 (100.0%)
P12		46 (52.9%)	41 (47.1%)	87 (100.0%)
P13		60 (68.2%)	28 (31.8%)	88 (100.0%)
P5		33 (40.2%)	49 (59.8%)	82 (100.0%)
P7		29 (60.4%)	19 (39.6%)	48 (100.0%)
P8		17 (33.3%)	34 (66.7%)	51 (100.0%)
P3		24 (44.4%)	30 (55.6%)	54 (100.0%)
P1		10 (40.0%)	15 (60.0%)	25 (100.0%)
P11		27 (56.2%)	21 (43.8%)	48 (100.0%)
Total		396 (53.4%)	346 (46.6%)	742 (100.0%)

#UTTERANCES by PARTICPANT facet TASK
gf_bar( PNUM ~., fill = ~ DATASET, data = df_coded) %>% 
  gf_facet_grid(.~TASK) + 
  labs(
    title = "Utterances by Participant, Dataset and Task",
    subtitle = "",
    x = "number of coded utterances",
    y = "participant",
    fill = "DATASET"
  )

2.4 by TIME

#DOTPLOT
ggplot(df_time, aes(x=rel_time, y = PNUM)) + 
  geom_point(alpha=0.5, size=3) +
  facet_grid(df_time$TASK) +
  scale_color_brewer(type="qual", palette = 3) +
  theme_minimal() + labs(
    title = "Participant Utterances over timecourse of Task",
    x= "timecourse of task (seconds)", y = "Participant",
    color = "Topic"
  ) 

#HISTOGRAMS BY TASK
ggplot(df_time, aes(x = rel_time)) + 
  geom_histogram(binwidth = 30,aes(y=..density..)) + 
  geom_density()+
  facet_grid(df_time$TASK) +
  theme_minimal() + labs(
    title = "Participant Utterances over timecourse of Task",
    x= "timecourse of task (seconds)", y = "frequency of utterances",
  ) + theme_minimal() + theme(legend.position = "blank")

## Warning: The dot-dot notation (`..density..`) was deprecated in ggplot2 3.4.0.
## ℹ Please use `after_stat(density)` instead.

3 TOP-CODE

3.1 by TASK

#COUNT BY TASK
ctable(x = df_coded$code_topic, 
       y = df_coded$TASK, 
       prop = "r")  

Cross-Tabulation, Row Proportions
code_topic * TASK
Data Frame: df_coded

	TASK	static	ixn	Total
code_topic
PROCESS		92 (57.5%)	68 (42.5%)	160 (100.0%)
DATASET		99 (56.2%)	77 (43.8%)	176 (100.0%)
VARIABLE		76 (62.3%)	46 (37.7%)	122 (100.0%)
RELATIONSHIP		129 (45.4%)	155 (54.6%)	284 (100.0%)
Total		396 (53.4%)	346 (46.6%)	742 (100.0%)

#DF SUMMARIZED BY TASK + DATASET
df_summary <- df_coded %>% 
  group_by(code_topic, TASK) %>% 
  dplyr::summarise(
    c = n()
  )

#STACKED BAR BY TASK
ggplot(df_summary, aes(x = TASK, y=c, fill= fct_rev(code_topic))) + 
  geom_col() + 
  geom_text(aes(label=c), size = 3, hjust = 0.5, vjust = 1.5, position = "stack") + 
  scale_fill_brewer(type="qual", palette = 3) +
  labs( title = "TOPICS by TASK",
        subtitle = "",
        x= "TASK", y = "count", fill="TOPIC") + theme_minimal() 

# + theme(legend.position = "blank")

3.2 by TASK and DATASET

#DF SUMMARIZED BY TASK + DATASET
df_summary <- df_coded %>%
  group_by(code_topic, TASK,DATASET) %>%
  dplyr::summarise(
    c = n()
  )

#STACKED BAR BY TASK FACET DATASET
ggplot(df_summary, aes(x = TASK, y=c, fill= fct_rev(code_topic))) +
  facet_wrap(df_summary$DATASET) +
  geom_col() +
  geom_text(aes(label=c), size = 3, hjust = 0.5, vjust = 1.5, position = "stack") +
  scale_fill_brewer(type="qual", palette = 3) +
  labs( title = "TOPICS by TASK and DATASET",
        subtitle = "",
        x= "TASK", y = "count", fill="TOPIC") + theme_minimal()

# + theme(legend.position = "blank")

3.3 by PARTICIPANT

#COUNT BY PARTICIPANT 
ctable(x = df_coded$PNUM, 
       y = df_coded$code_topic, 
       prop = "r")  

Cross-Tabulation, Row Proportions
PNUM * code_topic
Data Frame: df_coded

	code_topic	PROCESS	DATASET	VARIABLE	RELATIONSHIP	Total
PNUM
P6		6 (20.7%)	5 (17.2%)	5 (17.2%)	13 (44.8%)	29 (100.0%)
P9		19 (18.4%)	36 (35.0%)	23 (22.3%)	25 (24.3%)	103 (100.0%)
P10		8 (18.6%)	11 (25.6%)	8 (18.6%)	16 (37.2%)	43 (100.0%)
P2		3 (10.7%)	8 (28.6%)	16 (57.1%)	1 ( 3.6%)	28 (100.0%)
P4		10 (17.9%)	6 (10.7%)	11 (19.6%)	29 (51.8%)	56 (100.0%)
P12		21 (24.1%)	28 (32.2%)	17 (19.5%)	21 (24.1%)	87 (100.0%)
P13		41 (46.6%)	7 ( 8.0%)	8 ( 9.1%)	32 (36.4%)	88 (100.0%)
P5		6 ( 7.3%)	18 (22.0%)	18 (22.0%)	40 (48.8%)	82 (100.0%)
P7		14 (29.2%)	14 (29.2%)	7 (14.6%)	13 (27.1%)	48 (100.0%)
P8		10 (19.6%)	7 (13.7%)	3 ( 5.9%)	31 (60.8%)	51 (100.0%)
P3		7 (13.0%)	6 (11.1%)	2 ( 3.7%)	39 (72.2%)	54 (100.0%)
P1		8 (32.0%)	4 (16.0%)	0 ( 0.0%)	13 (52.0%)	25 (100.0%)
P11		7 (14.6%)	26 (54.2%)	4 ( 8.3%)	11 (22.9%)	48 (100.0%)
Total		160 (21.6%)	176 (23.7%)	122 (16.4%)	284 (38.3%)	742 (100.0%)

#TOPICS by PARTICPANT facet TASK
gf_bar( PNUM ~., fill = ~ fct_rev(code_topic), data = df_coded) %>% 
  gf_facet_grid(.~TASK) + 
  scale_fill_brewer(type="qual", palette = 3) +
  labs(
    title = "Utterances by Participant, Dataset and Task",
    subtitle = "",
    x = "number of coded utterances",
    y = "participant",
    fill = "TOPIC"
  ) + theme_minimal()

# #TOPICS by PARTICPANT facet TASK
# gf_bar( PNUM ~., fill = ~ fct_rev(code_topic), data = df_coded) %>% 
#   gf_facet_grid(DATASET~TASK) + 
#   scale_fill_brewer(type="qual", palette = 3) +
#   labs(
#     title = "Utterances by Participant, Dataset and Task",
#     subtitle = "",
#     x = "number of coded utterances",
#     y = "participant",
#     fill = "DATASET"
#   )

3.4 by TIME

#DOTPLOT
ggplot(df_time, aes(x=rel_time, y = PNUM, color=fct_rev(code_topic))) + 
  geom_point(alpha=0.5, size=3) +
  facet_grid(df_time$TASK) +
  scale_color_brewer(type="qual", palette = 3) +
  theme_minimal() + labs(
    title = "Topic of Utterances over timecourse of Task",
    x= "timecourse of task (seconds)", y = "Participant",
    color = "Topic"
  ) 

#HISTOGRAMS BY TASK
ggplot(df_time, aes(x = rel_time)) + 
  geom_histogram(binwidth = 30,aes(y=..density.., fill = fct_rev(code_topic), color = fct_rev(code_topic))) + 
  geom_density()+
  facet_grid(df_time$code_topic ~ df_time$TASK) +
  scale_fill_brewer(type="qual", palette = 3) +
  scale_color_brewer(type="qual", palette = 3) +
  theme_minimal() + labs(
    title = "Topic of Utterance over timecourse of Task",
    x= "timecourse of task (seconds)", y = "frequency of utterances",
    fill = "Topic"
  ) + theme_minimal() + theme(legend.position = "blank")

4 DETAIL—PROCESS

#PREP DATA FRAMES
df_process <- df_coded %>% 
  filter(code_topic=="PROCESS") %>% 
  dplyr::select(pid,PNUM,TASK,DATASET,code_detail)

df_time_process <- df_time %>% 
  filter(code_topic=="PROCESS") %>% 
  dplyr::select(pid,PNUM,TASK,DATASET,rel_time,code_detail)

df_summary <- df_process %>% 
  group_by(code_detail, TASK) %>% 
  dplyr::summarise(c = n())


#DETAILS BY TASK
ggplot(df_summary, aes(x = TASK, y=c, fill= code_detail)) + 
  geom_col() + 
  geom_text(aes(label=c), size = 3, hjust = 0.5, vjust = 1.5, position = "stack") + 
  scale_fill_brewer(palette="Set2") +
  labs( title = "PROCESS Utterances by TASK",
        subtitle = "",
        x= "TASK", y = "count") + theme_minimal() 

#DETAILS DOTPLOT
ggplot(df_time_process, aes(x=rel_time, y = PNUM, color=fct_rev(code_detail))) + 
  geom_point(alpha=0.5, size=3) +
  facet_grid(df_time_process$TASK) +
  scale_color_brewer(palette="Set2") +
  theme_minimal() + labs(
    title = "PROCESS Utterances by timecourse of Task",
    x= "timecourse of task (seconds)", y = "Participant",
    color = "Topic"
  ) 

#DETAIL HISTOGRAMS BY TASK
ggplot(df_time_process, aes(x = rel_time, fill = fct_rev(code_detail))) + 
  geom_histogram(binwidth = 30) + 
  facet_grid(df_time_process$TASK ~ df_time_process$code_detail ) +
  scale_fill_brewer(palette = "Set2") +
  theme_minimal() + labs(
    title = "PROCESS Utterances by timecourse of Task",
    x= "timecourse of task (seconds)", y = "frequency of utterances"
  ) + theme_minimal() + theme(legend.position = "blank")

5 DETAIL—DATASET

#PREP DATA FRAMES
df_dataset <- df_coded %>% 
  filter(code_topic=="DATASET") %>% 
  dplyr::select(pid,PNUM,TASK,DATASET,code_detail)

df_time_dataset <- df_time %>% 
  filter(code_topic=="DATASET") %>% 
  dplyr::select(pid,PNUM,TASK,DATASET,rel_time,code_detail)

df_summary <- df_dataset %>% 
  group_by(code_detail, TASK) %>% 
  dplyr::summarise(c = n())


#DETAILS BY TASK
ggplot(df_summary, aes(x = TASK, y=c, fill= code_detail)) + 
  geom_col() + 
  geom_text(aes(label=c), size = 3, hjust = 0.5, vjust = 1.5, position = "stack") + 
  scale_fill_brewer(type = "seq", palette = 4) +
  labs( title = "DATASET Utterances by TASK",
        subtitle = "",
        x= "TASK", y = "count") + theme_minimal() 

#DETAILS DOTPLOT
ggplot(df_time_dataset, aes(x=rel_time, y = PNUM, color=fct_rev(code_detail))) + 
  geom_point(alpha=0.5, size=3) +
  facet_grid(df_time_dataset$TASK) +
  scale_color_brewer(type = "seq", palette = 4) +
  theme_minimal() + labs(
    title = "VARIABLE Utterances by timecourse of Task",
    x= "timecourse of task (seconds)", y = "Participant",
    color = "Topic"
  ) 

#DETAIL HISTOGRAMS BY TASK
ggplot(df_time_dataset, aes(x = rel_time, fill = fct_rev(code_detail))) + 
  geom_histogram(binwidth = 30) + 
  facet_grid(df_time_dataset$TASK ~ df_time_dataset$code_detail ) +
  scale_fill_brewer(type = "seq", palette = 4) +
  theme_minimal() + labs(
    title = "VARIABLE Utterances by timecourse of Task",
    x= "timecourse of task (seconds)", y = "frequency of utterances"
  ) + theme_minimal() + theme(legend.position = "blank")

6 DETAIL—VARIABLE

#PREP DATA FRAMES
df_variable <- df_coded %>% 
  filter(code_topic=="VARIABLE") %>% 
  dplyr::select(pid,PNUM,TASK,DATASET,code_detail)

df_time_variable <- df_time %>% 
  filter(code_topic=="VARIABLE") %>% 
  dplyr::select(pid,PNUM,TASK,DATASET,rel_time,code_detail)

df_summary <- df_variable %>% 
  group_by(code_detail, TASK) %>% 
  dplyr::summarise(c = n())


#DETAILS BY TASK
ggplot(df_summary, aes(x = TASK, y=c, fill= code_detail)) + 
  geom_col() + 
  geom_text(aes(label=c), size = 3, hjust = 0.5, vjust = 1.5, position = "stack") + 
  scale_fill_brewer(type = "seq", palette = 5) +
  labs( title = "VARIABLE Utterances by TASK",
        subtitle = "",
        x= "TASK", y = "count") + theme_minimal() 

#DETAILS DOTPLOT
ggplot(df_time_variable, aes(x=rel_time, y = PNUM, color=fct_rev(code_detail))) + 
  geom_point(alpha=0.5, size=3) +
  facet_grid(df_time_variable$TASK) +
  scale_color_brewer(type = "seq", palette = 5) +
  theme_minimal() + labs(
    title = "VARIABLE Utterances by timecourse of Task",
    x= "timecourse of task (seconds)", y = "Participant",
    color = "Topic"
  ) 

#DETAIL HISTOGRAMS BY TASK
ggplot(df_time_variable, aes(x = rel_time, fill = fct_rev(code_detail))) + 
  geom_histogram(binwidth = 30) + 
  facet_grid(df_time_variable$TASK ~ df_time_variable$code_detail) +
  scale_fill_brewer(type = "seq", palette = 5) +
  theme_minimal() + labs(
    title = "VARIABLE Utterances by timecourse of Task",
    x= "timecourse of task (seconds)", y = "frequency of utterances"
  ) + theme_minimal() + theme(legend.position = "blank")

7 DETAIL—RELATIONSHIP

#PREP DATA FRAMES
df_relationship <- df_coded %>% 
  filter(code_topic=="RELATIONSHIP") %>% 
  dplyr::select(pid,PNUM,TASK,DATASET,code_detail)

df_time_relationship <- df_time %>% 
  filter(code_topic=="RELATIONSHIP") %>% 
  dplyr::select(pid,PNUM,TASK,DATASET,rel_time,code_detail)

df_summary <- df_relationship %>% 
  group_by(code_detail, TASK) %>% 
  dplyr::summarise(c = n())


#DETAILS BY TASK
ggplot(df_summary, aes(x = TASK, y=c, fill= code_detail)) + 
  geom_col() + 
  geom_text(aes(label=c), size = 3, hjust = 0.5, vjust = 1.5, position = "stack") + 
  scale_fill_brewer(type = "seq", palette = 3) +
  labs( title = "RELATIONSHIP Utterances by TASK",
        subtitle = "",
        x= "TASK", y = "count") + theme_minimal() 

#DETAILS DOTPLOT
ggplot(df_time_relationship, aes(x=rel_time, y = PNUM, color=fct_rev(code_detail))) + 
  geom_point(alpha=0.5, size=3) +
  facet_grid(df_time_relationship$TASK) +
  scale_color_brewer(type = "seq", palette = 3) +
  theme_minimal() + labs(
    title = "RELATIONSHIP Utterances by timecourse of Task",
    x= "timecourse of task (seconds)", y = "Participant",
    color = "Topic"
  ) 

#DETAIL HISTOGRAMS BY TASK
ggplot(df_time_relationship, aes(x = rel_time, fill = fct_rev(code_detail))) + 
  geom_histogram(binwidth = 30) + 
  facet_grid(df_time_relationship$TASK ~ df_time_relationship$code_detail) +
  scale_fill_brewer(type = "seq", palette = 3) +
  theme_minimal() + labs(
    title = "RELATIONSHIP Utterances by timecourse of Task",
    x= "timecourse of task (seconds)", y = "frequency of utterances"
  ) + theme_minimal() + theme(legend.position = "blank")

8 REPRESENTATIONS

8.1 Representation Groups

How many representations were created?

print("BY TASK")

[1] “BY TASK”

ctable(df_coded$reps_group, df_coded$TASK,
    prop = "t",
    plain.ascii = FALSE)

8.1.1 Cross-Tabulation, Total Proportions

8.1.1.1 reps_group * TASK

Data Frame: df_coded

	TASK	static	ixn	Total
reps_group
barplot		8 ( 1.1%)	8 ( 1.1%)	16 ( 2.2%)
columns		4 ( 0.5%)	0 ( 0.0%)	4 ( 0.5%)
columns_data_dictionary		1 ( 0.1%)	0 ( 0.0%)	1 ( 0.1%)
data_dictionary		29 ( 3.9%)	30 ( 4.0%)	59 ( 8.0%)
data_dictionary_dataframe		1 ( 0.1%)	0 ( 0.0%)	1 ( 0.1%)
data_dictionary_describe		9 ( 1.2%)	1 ( 0.1%)	10 ( 1.3%)
dataframe		55 ( 7.4%)	25 ( 3.4%)	80 ( 10.8%)
dataframe_describe		1 ( 0.1%)	0 ( 0.0%)	1 ( 0.1%)
dataframe_heatmap		0 ( 0.0%)	1 ( 0.1%)	1 ( 0.1%)
dataframe_pairplot		1 ( 0.1%)	0 ( 0.0%)	1 ( 0.1%)
dataframe_profile		1 ( 0.1%)	2 ( 0.3%)	3 ( 0.4%)
describe		11 ( 1.5%)	13 ( 1.8%)	24 ( 3.2%)
describe_profile		1 ( 0.1%)	0 ( 0.0%)	1 ( 0.1%)
double-profiler		13 ( 1.8%)	8 ( 1.1%)	21 ( 2.8%)
heatmap		13 ( 1.8%)	7 ( 0.9%)	20 ( 2.7%)
hist		6 ( 0.8%)	0 ( 0.0%)	6 ( 0.8%)
info		9 ( 1.2%)	6 ( 0.8%)	15 ( 2.0%)
lineplot		38 ( 5.1%)	0 ( 0.0%)	38 ( 5.1%)
Multi-view Chart		9 ( 1.2%)	50 ( 6.7%)	59 ( 8.0%)
none		17 ( 2.3%)	14 ( 1.9%)	31 ( 4.2%)
pairplot		38 ( 5.1%)	14 ( 1.9%)	52 ( 7.0%)
profile		47 ( 6.3%)	59 ( 8.0%)	106 ( 14.3%)
python		30 ( 4.0%)	30 ( 4.0%)	60 ( 8.1%)
scatterplot		48 ( 6.5%)	78 (10.5%)	126 ( 17.0%)
stripplot		6 ( 0.8%)	0 ( 0.0%)	6 ( 0.8%)
Total		396 (53.4%)	346 (46.6%)	742 (100.0%)

print("BY DATASET")

[1] “BY DATASET”

ctable(df_coded$reps_group, df_coded$DATASET,
    prop = "t",
    plain.ascii = FALSE)

8.1.2 Cross-Tabulation, Total Proportions

8.1.2.1 reps_group * DATASET

Data Frame: df_coded

	DATASET	happiness	space	Total
reps_group
barplot		3 ( 0.4%)	13 ( 1.8%)	16 ( 2.2%)
columns		1 ( 0.1%)	3 ( 0.4%)	4 ( 0.5%)
columns_data_dictionary		0 ( 0.0%)	1 ( 0.1%)	1 ( 0.1%)
data_dictionary		24 ( 3.2%)	35 ( 4.7%)	59 ( 8.0%)
data_dictionary_dataframe		1 ( 0.1%)	0 ( 0.0%)	1 ( 0.1%)
data_dictionary_describe		7 ( 0.9%)	3 ( 0.4%)	10 ( 1.3%)
dataframe		37 ( 5.0%)	43 ( 5.8%)	80 ( 10.8%)
dataframe_describe		0 ( 0.0%)	1 ( 0.1%)	1 ( 0.1%)
dataframe_heatmap		1 ( 0.1%)	0 ( 0.0%)	1 ( 0.1%)
dataframe_pairplot		1 ( 0.1%)	0 ( 0.0%)	1 ( 0.1%)
dataframe_profile		0 ( 0.0%)	3 ( 0.4%)	3 ( 0.4%)
describe		16 ( 2.2%)	8 ( 1.1%)	24 ( 3.2%)
describe_profile		0 ( 0.0%)	1 ( 0.1%)	1 ( 0.1%)
double-profiler		3 ( 0.4%)	18 ( 2.4%)	21 ( 2.8%)
heatmap		15 ( 2.0%)	5 ( 0.7%)	20 ( 2.7%)
hist		0 ( 0.0%)	6 ( 0.8%)	6 ( 0.8%)
info		9 ( 1.2%)	6 ( 0.8%)	15 ( 2.0%)
lineplot		38 ( 5.1%)	0 ( 0.0%)	38 ( 5.1%)
Multi-view Chart		38 ( 5.1%)	21 ( 2.8%)	59 ( 8.0%)
none		20 ( 2.7%)	11 ( 1.5%)	31 ( 4.2%)
pairplot		44 ( 5.9%)	8 ( 1.1%)	52 ( 7.0%)
profile		55 ( 7.4%)	51 ( 6.9%)	106 ( 14.3%)
python		9 ( 1.2%)	51 ( 6.9%)	60 ( 8.1%)
scatterplot		102 (13.7%)	24 ( 3.2%)	126 ( 17.0%)
stripplot		0 ( 0.0%)	6 ( 0.8%)	6 ( 0.8%)
Total		424 (57.1%)	318 (42.9%)	742 (100.0%)

8.2 TOPICS + REPRESENTATIONS

df <- df_coded 

ctable(df$reps_group, df$code_topic,
    prop = "t",
    plain.ascii = FALSE)

8.2.1 Cross-Tabulation, Total Proportions

8.2.1.1 reps_group * code_topic

Data Frame: df

	code_topic	PROCESS	DATASET	VARIABLE	RELATIONSHIP	Total
reps_group
barplot		5 ( 0.7%)	1 ( 0.1%)	2 ( 0.3%)	8 ( 1.1%)	16 ( 2.2%)
columns		0 ( 0.0%)	2 ( 0.3%)	1 ( 0.1%)	1 ( 0.1%)	4 ( 0.5%)
columns_data_dictionary		0 ( 0.0%)	1 ( 0.1%)	0 ( 0.0%)	0 ( 0.0%)	1 ( 0.1%)
data_dictionary		7 ( 0.9%)	41 ( 5.5%)	2 ( 0.3%)	9 ( 1.2%)	59 ( 8.0%)
data_dictionary_dataframe		0 ( 0.0%)	1 ( 0.1%)	0 ( 0.0%)	0 ( 0.0%)	1 ( 0.1%)
data_dictionary_describe		0 ( 0.0%)	1 ( 0.1%)	7 ( 0.9%)	2 ( 0.3%)	10 ( 1.3%)
dataframe		21 ( 2.8%)	34 ( 4.6%)	7 ( 0.9%)	18 ( 2.4%)	80 ( 10.8%)
dataframe_describe		0 ( 0.0%)	1 ( 0.1%)	0 ( 0.0%)	0 ( 0.0%)	1 ( 0.1%)
dataframe_heatmap		0 ( 0.0%)	0 ( 0.0%)	0 ( 0.0%)	1 ( 0.1%)	1 ( 0.1%)
dataframe_pairplot		0 ( 0.0%)	0 ( 0.0%)	1 ( 0.1%)	0 ( 0.0%)	1 ( 0.1%)
dataframe_profile		1 ( 0.1%)	2 ( 0.3%)	0 ( 0.0%)	0 ( 0.0%)	3 ( 0.4%)
describe		2 ( 0.3%)	14 ( 1.9%)	3 ( 0.4%)	5 ( 0.7%)	24 ( 3.2%)
describe_profile		0 ( 0.0%)	0 ( 0.0%)	1 ( 0.1%)	0 ( 0.0%)	1 ( 0.1%)
double-profiler		2 ( 0.3%)	4 ( 0.5%)	3 ( 0.4%)	12 ( 1.6%)	21 ( 2.8%)
heatmap		5 ( 0.7%)	0 ( 0.0%)	0 ( 0.0%)	15 ( 2.0%)	20 ( 2.7%)
hist		0 ( 0.0%)	0 ( 0.0%)	3 ( 0.4%)	3 ( 0.4%)	6 ( 0.8%)
info		5 ( 0.7%)	8 ( 1.1%)	1 ( 0.1%)	1 ( 0.1%)	15 ( 2.0%)
lineplot		20 ( 2.7%)	0 ( 0.0%)	0 ( 0.0%)	18 ( 2.4%)	38 ( 5.1%)
Multi-view Chart		11 ( 1.5%)	10 ( 1.3%)	13 ( 1.8%)	25 ( 3.4%)	59 ( 8.0%)
none		16 ( 2.2%)	5 ( 0.7%)	2 ( 0.3%)	8 ( 1.1%)	31 ( 4.2%)
pairplot		10 ( 1.3%)	1 ( 0.1%)	10 ( 1.3%)	31 ( 4.2%)	52 ( 7.0%)
profile		21 ( 2.8%)	20 ( 2.7%)	50 ( 6.7%)	15 ( 2.0%)	106 ( 14.3%)
python		7 ( 0.9%)	22 ( 3.0%)	11 ( 1.5%)	20 ( 2.7%)	60 ( 8.1%)
scatterplot		27 ( 3.6%)	8 ( 1.1%)	5 ( 0.7%)	86 (11.6%)	126 ( 17.0%)
stripplot		0 ( 0.0%)	0 ( 0.0%)	0 ( 0.0%)	6 ( 0.8%)	6 ( 0.8%)
Total		160 (21.6%)	176 (23.7%)	122 (16.4%)	284 (38.3%)	742 (100.0%)

8.3 TODO PICK UP HERE

9 MODELLING

#DEFINE DATAFRAME
df <- df_coded %>% select(pid, uid, TASK, DATASET) 
  
#MOSAIC PLOT
mosaic(formula = ~DATASET + TASK, 
       data = df,
       main = "Proportion of Utterances by TASK and DATASET", 
       sub = "u = 734 utterance-codes",
       labeling = labeling_values,
       labeling_args = list(set_varnames = c(graph = "TASK",
                            datset = "DATASET")))

9.1 UTTERANCES

How much variance in number of utterances is explained DATASET, TASK and PARTICIPANT?

9.1.1 OLS Mixed Effects Models

#DEFINE DATAFRAME
df <- df_coded %>% group_by(pid, DATASET, TASK) %>% 
  dplyr::summarise( .groups = "keep",
    n_utterances = n()
  )

#NUMBER UTTERANCES predicted by DATASET + TASK | participatnt--> MIXED LINEAR REGRESSION
print("LMER, UTTERANCES ~ DATASET + TASK")

## [1] "LMER, UTTERANCES ~ DATASET + TASK"

mm1 <- lmer(n_utterances ~ DATASET + TASK+ (1|pid), data = df)
paste("Model")

## [1] "Model"

summ(mm1)

Observations	26
Dependent variable	n_utterances
Type	Mixed effects linear regression

AIC	199.41
BIC	205.70
Pseudo-R² (fixed effects)	0.09
Pseudo-R² (total)	0.64

Fixed Effects
	Est.	S.E.	t val.	d.f.	p
(Intercept)	34.11	4.23	8.07	20.73	0.00
DATASETspace	-7.90	3.48	-2.27	11.00	0.04
TASKixn	-3.24	3.48	-0.93	11.00	0.37
p values calculated using Satterthwaite d.f.

Random Effects
Group	Parameter	Std. Dev.
pid	(Intercept)	10.97
Residual		8.85

Grouping Variables
Group	# groups	ICC
pid	13	0.61

paste("Partition Variance")

## [1] "Partition Variance"

anova(mm1)

## Type III Analysis of Variance Table with Satterthwaite's method
##         Sum Sq Mean Sq NumDF DenDF F value  Pr(>F)  
## DATASET 403.75  403.75     1    11  5.1577 0.04423 *
## TASK     67.75   67.75     1    11  0.8655 0.37218  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

paste("Confidence Interval on Parameter Estimates")

## [1] "Confidence Interval on Parameter Estimates"

confint(mm1)

## Computing profile confidence intervals ...

##                   2.5 %    97.5 %
## .sig01         5.404680 17.974025
## .sigma         5.787943 12.644890
## (Intercept)   25.894875 42.324905
## DATASETspace -14.673480 -1.136044
## TASKixn      -10.006813  3.530623

report(mm1) #sanity check

## We fitted a linear mixed model (estimated using REML and nloptwrap optimizer)
## to predict n_utterances with DATASET and TASK (formula: n_utterances ~ DATASET
## + TASK). The model included pid as random effect (formula: ~1 | pid). The
## model's total explanatory power is substantial (conditional R2 = 0.64) and the
## part related to the fixed effects alone (marginal R2) is of 0.09. The model's
## intercept, corresponding to DATASET = happiness and TASK = static, is at 34.11
## (95% CI [25.32, 42.90], t(21) = 8.07, p < .001). Within this model:
## 
##   - The effect of DATASET [space] is statistically significant and negative (beta
## = -7.90, 95% CI [-15.14, -0.67], t(21) = -2.27, p = 0.034; Std. beta = -0.55,
## 95% CI [-1.05, -0.05])
##   - The effect of TASK [ixn] is statistically non-significant and negative (beta
## = -3.24, 95% CI [-10.48, 4.00], t(21) = -0.93, p = 0.363; Std. beta = -0.22,
## 95% CI [-0.73, 0.28])
## 
## Standardized parameters were obtained by fitting the model on a standardized
## version of the dataset. 95% Confidence Intervals (CIs) and p-values were
## computed using a Wald t-distribution approximation.

plot_model(mm1,  show.intercept = TRUE)

check_model(mm1)

#NUMBER UTTERANCES predicted by DATASET * TASK  | participatnt--> MIXED LINEAR REGRESSION
print("LMER, UTTERANCES ~ DATASET X TASK")

## [1] "LMER, UTTERANCES ~ DATASET X TASK"

mm2 <- lmer(n_utterances ~ DATASET * TASK + (1|pid), data = df)
paste("Model")

## [1] "Model"

summ(mm2)

Observations	26
Dependent variable	n_utterances
Type	Mixed effects linear regression

AIC	193.70
BIC	201.25
Pseudo-R² (fixed effects)	0.12
Pseudo-R² (total)	0.66

Fixed Effects
	Est.	S.E.	t val.	d.f.	p
(Intercept)	36.57	5.41	6.76	15.92	0.00
DATASETspace	-13.24	7.96	-1.66	15.92	0.12
TASKixn	-8.57	7.96	-1.08	15.92	0.30
DATASETspace:TASKixn	10.67	14.32	0.74	11.00	0.47
p values calculated using Satterthwaite d.f.

Random Effects
Group	Parameter	Std. Dev.
pid	(Intercept)	11.25
Residual		8.85

Grouping Variables
Group	# groups	ICC
pid	13	0.62

paste("Partition Variance")

## [1] "Partition Variance"

anova(mm2)

## Type III Analysis of Variance Table with Satterthwaite's method
##              Sum Sq Mean Sq NumDF DenDF F value  Pr(>F)  
## DATASET      403.75  403.75     1    11  5.1577 0.04423 *
## TASK          67.75   67.75     1    11  0.8655 0.37218  
## DATASET:TASK  43.44   43.44     1    11  0.5549 0.47195  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

paste("Confidence Interval on Parameter Estimates")

## [1] "Confidence Interval on Parameter Estimates"

confint(mm2)

## Computing profile confidence intervals ...

##                           2.5 %    97.5 %
## .sig01                 4.993856 17.494032
## .sigma                 5.787971 12.644901
## (Intercept)           26.281282 46.861575
## DATASETspace         -28.384780  1.908591
## TASKixn              -23.718114  6.575257
## DATASETspace:TASKixn -17.180459 38.513793

report(mm2) #sanity check

## We fitted a linear mixed model (estimated using REML and nloptwrap optimizer)
## to predict n_utterances with DATASET and TASK (formula: n_utterances ~ DATASET
## * TASK). The model included pid as random effect (formula: ~1 | pid). The
## model's total explanatory power is substantial (conditional R2 = 0.66) and the
## part related to the fixed effects alone (marginal R2) is of 0.12. The model's
## intercept, corresponding to DATASET = happiness and TASK = static, is at 36.57
## (95% CI [25.29, 47.85], t(20) = 6.76, p < .001). Within this model:
## 
##   - The effect of DATASET [space] is statistically non-significant and negative
## (beta = -13.24, 95% CI [-29.84, 3.37], t(20) = -1.66, p = 0.112; Std. beta =
## -0.92, 95% CI [-2.07, 0.23])
##   - The effect of TASK [ixn] is statistically non-significant and negative (beta
## = -8.57, 95% CI [-25.18, 8.04], t(20) = -1.08, p = 0.294; Std. beta = -0.59,
## 95% CI [-1.75, 0.56])
##   - The effect of DATASET [space] × TASK [ixn] is statistically non-significant
## and positive (beta = 10.67, 95% CI [-19.20, 40.54], t(20) = 0.74, p = 0.465;
## Std. beta = 0.74, 95% CI [-1.33, 2.81])
## 
## Standardized parameters were obtained by fitting the model on a standardized
## version of the dataset. 95% Confidence Intervals (CIs) and p-values were
## computed using a Wald t-distribution approximation.

plot_model(mm2,  show.intercept = TRUE)

check_model(mm2)

9.1.2 POISSON Mixed Effects Models

# 
# #NUMBER UTTERANCES predicted by TASK + DATASET  | participatnt--> POISSON MIXED LINEAR REGRESSION
# print("POISSON-MER, UTTERANCES ~ DATASET + TASK")
# pmm1 <- glmer(n_utterances ~ TASK + DATASET + (1|pid), data = df, family = "poisson")
# paste("Model")
# summ(pmm1)
# paste("Partition Variance")
# anova(pmm1)
# paste("Confidence Interval on Parameter Estimates")
# confint(pmm1)
# report(pmm1) #sanity check
# plot_model(pmm1,  show.intercept = TRUE)
# check_model(pmm1)
# 
# #NUMBER UTTERANCES predicted by TASK X DATASET  | participatnt--> POISSON MIXED LINEAR REGRESSION
# print("POISSON-MER, UTTERANCES ~ DATASET X TASK")
# pmm2 <- glmer(n_utterances ~ TASK * DATASET + (1|pid), data = df, family = "poisson")
# paste("Model")
# summ(pmm2)
# paste("Partition Variance")
# anova(pmm2)
# paste("Confidence Interval on Parameter Estimates")
# confint(pmm2)
# report(pmm2) #sanity check
# plot_model(pmm2,  show.intercept = TRUE)
# check_model(pmm2)